Conventional method for detecting obstacles for autonomous mobile vehicle uses various sensors such as CCD, 2D laser radar, etc. It also determines every detected object, which is not planar, is regarded as an obstacle (Documents 1 and 2) based on the assumption that the operating environment is relatively planar. This method of detecting obstacles is appropriate for use indoors or in the city environment, where the amount of data to be processed is small and the calculation time for the data is relatively short.
In non-paved roads or field areas, the conventional method generates a 3 dimensional World Model including elevation information by using sensors (CCD, IR, laser radar, radar, etc.). This method also detects an obstacle by analyzing drivability based on the World Model (Documents 3 and 4). In Addition, this method requires much calculation time since a lot of data from various sensors should be obtained and analyzed.
Additionally, in the case of application where detecting and avoiding obstacles that emerge suddenly at a short distance, obstacles are detected by using one or more of 2D laser radars and avoidance is carried out for a fast response on the assumption that the driving environment is relatively planar (Documents 5 and 6).
The conventional methods, however, are not appropriate for the autonomous mobile vehicle such as armored vehicle which performs an abrupt movement on a non-paved road or field area since only one 2D laser radar built in the autonomous mobile vehicle, or a plurality of 2D laser radars arranged in parallel with the base plane of the vehicle is used. When lateral or longitudinal force acts on the vehicle due to an abrupt change of direction or speed, the autonomous vehicle is slanted by the lateral or longitudinal force in the right or left direction (roll direction), or in the front or rear direction (pitch direction). In this case, a relatively planar ground is erroneously recognized as an obstacle.
FIG. 5 illustrates the radiation pattern of beam from the 2D laser radars, which are arranged in parallel with the base plane of the autonomous vehicle when the right and left suspension devices are pressed differently due to an abrupt change of direction of the vehicle. It can be seen from the figure that the distance data between left and right direction are different although the ground is planar when the autonomous vehicle is inclined in the roll direction. As a result, planar ground can be erroneously recognized as an obstacle when the autonomous vehicle is inclined in the roll direction. Therefore, the vehicle will make an avoidance action although there is no obstacle.